China's Zhipu AI Launches Open-Source GLM-4.5 Leading AI Model Amid Global Buzz
On July 31, 2025, Zhipu AI, a leading Chinese artificial intelligence firm, officially released its latest flagship model, the GLM-4.5. This new large language model quickly garnered international media coverage for its impressive performance and full open-source strategy. Positioned as a foundational model for various applications, the GLM-4.5 showcases remarkable capabilities in complex reasoning, code generation, and effective agent interaction. These advancements place it among the top contenders in the global AI landscape.
In an exclusive interview with the Global Times, Zhipu's CEO, Zhang Peng, elaborated on the unique features of GLM-4.5, including its ability to perform exceptionally well across an array of tasks. The model’s design allows it to deliver high precision in full-stack development tasks, such as creating complex applications and interactive web pages. For instance, developers can utilize GLM-4.5 to build functional search engines or interactive gaming experiences, demonstrating its versatility in real-world applications.
Zhipu's new model has achieved state-of-the-art (SOTA) performance in several benchmarks that assess AI capabilities comprehensively. According to Zhipu, GLM-4.5 ranks third globally and first among its Chinese competitors and open-source models in terms of average composite scores. Notably, it is engineered for efficiency, boasting costs as low as 0.8 yuan ($0.11) per million input tokens, which is significantly lower than many mainstream AI models. This cost-effective approach, combined with its ability to process over 100 tokens per second, positions GLM-4.5 as a high-performing choice in the increasingly competitive AI market.
The GLM-4.5 is built on an innovative architecture termed ‘agentic AI’, which disassembles tasks into manageable units, boosting accuracy during processing. Additionally, this model only requires eight NVIDIA H20 chips to operate, half the hardware needed by competing models, further reinforcing Zhipu's commitment to making advanced AI technology more accessible.
One of the model's most groundbreaking aspects is its full open-source release under the MIT license. This decision reflects Zhipu's belief in the power of collaborative development and its desire to contribute to the academic and professional AI community. Open-sourcing GLM-4.5 is expected to stimulate dialogue in both industry and academia regarding AI advancements and the future of artificial general intelligence (AGI). Zhang highlighted that this release encourages innovative thinking, particularly in the context of generational shifts in AI developments, as more companies engage with the opportunities presented by open-source models.
International digital platforms and media quickly recognized the significance of GLM-4.5 after its launch. Within hours, the model featured prominently on social media and AI discourse forums, reflecting a growing interest in China’s advancements in AI technology. Reporters from major tech outlets like CNBC and Techi noted that Zhipu’s strategic move potentially sets a new standard for competitors globally, leveraging lower costs and higher accessibility.
As a subsidiary of Tsinghua University, one of China's foremost educational institutions, Zhipu AI was founded with the intent to lead in AI innovations. Although it started its journey later than some U.S. counterparts, it has rapidly emerged as a significant player in the AI field. Zhang remarks that like OpenAI, which also pioneered large AI models, Zhipu's mission is to broaden the capabilities of its models, but with an open-source ethos that is increasingly rare in a sector dominated by closed systems.
The burgeoning development sector in China reflects not only a technological race but also a response to global pressures and market dynamics. With recent reports highlighting that China has released 1,509 large AI models—the largest share worldwide—the country is positioning itself as an important contributor to the global tech scene. Zhipu's GLM-4.5 represents a crucial step in solidifying China’s role in the AI landscape amidst complex international competition.
Zhang envisions the future of Zhipu as a continued commitment to exploring the frontiers of intelligence, engaging with global scientists and developers to enhance AI's capabilities and push for breakthroughs in understanding AGI. As AI technology continues to evolve, Zhipu aims to remain at the forefront, contributing to a framework that supports both innovation and practical applications in an increasingly AI-driven world.
In an exclusive interview with the Global Times, Zhipu's CEO, Zhang Peng, elaborated on the unique features of GLM-4.5, including its ability to perform exceptionally well across an array of tasks. The model’s design allows it to deliver high precision in full-stack development tasks, such as creating complex applications and interactive web pages. For instance, developers can utilize GLM-4.5 to build functional search engines or interactive gaming experiences, demonstrating its versatility in real-world applications.
Zhipu's new model has achieved state-of-the-art (SOTA) performance in several benchmarks that assess AI capabilities comprehensively. According to Zhipu, GLM-4.5 ranks third globally and first among its Chinese competitors and open-source models in terms of average composite scores. Notably, it is engineered for efficiency, boasting costs as low as 0.8 yuan ($0.11) per million input tokens, which is significantly lower than many mainstream AI models. This cost-effective approach, combined with its ability to process over 100 tokens per second, positions GLM-4.5 as a high-performing choice in the increasingly competitive AI market.
The GLM-4.5 is built on an innovative architecture termed ‘agentic AI’, which disassembles tasks into manageable units, boosting accuracy during processing. Additionally, this model only requires eight NVIDIA H20 chips to operate, half the hardware needed by competing models, further reinforcing Zhipu's commitment to making advanced AI technology more accessible.
One of the model's most groundbreaking aspects is its full open-source release under the MIT license. This decision reflects Zhipu's belief in the power of collaborative development and its desire to contribute to the academic and professional AI community. Open-sourcing GLM-4.5 is expected to stimulate dialogue in both industry and academia regarding AI advancements and the future of artificial general intelligence (AGI). Zhang highlighted that this release encourages innovative thinking, particularly in the context of generational shifts in AI developments, as more companies engage with the opportunities presented by open-source models.
International digital platforms and media quickly recognized the significance of GLM-4.5 after its launch. Within hours, the model featured prominently on social media and AI discourse forums, reflecting a growing interest in China’s advancements in AI technology. Reporters from major tech outlets like CNBC and Techi noted that Zhipu’s strategic move potentially sets a new standard for competitors globally, leveraging lower costs and higher accessibility.
As a subsidiary of Tsinghua University, one of China's foremost educational institutions, Zhipu AI was founded with the intent to lead in AI innovations. Although it started its journey later than some U.S. counterparts, it has rapidly emerged as a significant player in the AI field. Zhang remarks that like OpenAI, which also pioneered large AI models, Zhipu's mission is to broaden the capabilities of its models, but with an open-source ethos that is increasingly rare in a sector dominated by closed systems.
The burgeoning development sector in China reflects not only a technological race but also a response to global pressures and market dynamics. With recent reports highlighting that China has released 1,509 large AI models—the largest share worldwide—the country is positioning itself as an important contributor to the global tech scene. Zhipu's GLM-4.5 represents a crucial step in solidifying China’s role in the AI landscape amidst complex international competition.
Zhang envisions the future of Zhipu as a continued commitment to exploring the frontiers of intelligence, engaging with global scientists and developers to enhance AI's capabilities and push for breakthroughs in understanding AGI. As AI technology continues to evolve, Zhipu aims to remain at the forefront, contributing to a framework that supports both innovation and practical applications in an increasingly AI-driven world.
Topics Consumer Technology）
【About Using Articles】
You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.
【About Links】
Links are free to use.